Font Size: a A A

Research On Improved Particle Swarm Optimization For Single Objective Problem

Posted on:2019-06-03Degree:MasterType:Thesis
Country:ChinaCandidate:C Q LiFull Text:PDF
GTID:2428330545974359Subject:Mathematics
Abstract/Summary:PDF Full Text Request
Particle swarm optimization(PSO)is a kind of random search algorithm based on swarm by simulating bird foraging behavior.PSO can effectively solve the nonlinear and complex optimization problems which the traditional algorithms have no ability to solve.In the evolution process,PSO firstly initializes a group of initial solutions randomly,and then the positions of the particles are updated by constantly iterating and replacing.Finally,the process can make the population evolve towards to the optimal location.As PSO is easy to operate and realize,it has received the extensive attention in the international and domestic scholars.However,PSO also has the premature convergence like other evolutionary algorithms.In view of this,the main research work of this paper is as follows:(1)Analysis of particle trajectories and convergence of PSOPSO is a random search algorithm whose evolution equation is a discrete dynamic system.By analyzing the state change of dynamic system,determining the necessary and sufficient condition for algorithm convergence which can provide theoretical guarantee for algorithm improvement.(2)PSO based on the neighborhood fitness value(NFPSO).We analyzed the neighborhood structure of the standard PSO,and explored the influence of neighbor structure on the algorithm.The strategy of the neighborhood fitness value is proposed to make full use of individual information.The strategy can improve the diversity of population effectively.(3)Hybrid learning PSO based on neighborhood adaptive value strategy and comprehensive learning strategy(NCPSO).By learning from the strategy of the standard PSO that jumped out of the local optimal solution and improved the diversity of the population,we proposed a mixed learning strategy to ensure the efficiency of the algorithm.(4)Hybrid learning particle swarm algorithm(NDPSO)based on neighborhood fitness strategy and dynamic tournament strategy.Based on the advantages of thehybrid strategy,we borrow the advantages of sample selection from the championship strategy that make full use of all individual information to enhance the ability of algorithm optimization.
Keywords/Search Tags:Particle swarm optimization, Premature convergence, Topological structure, Comprehensive learning strategy, Dynamic tournament strategy, Mixture learning strategy
PDF Full Text Request
Related items